📄 中文摘要
红树林对于缓解气候变化至关重要，需要可靠的监测以实现有效保护。深度学习已成为红树林检测的强大工具，但其进展受现有数据集限制。具体而言，许多现有资源仅提供年度地图产品，缺乏精心策划的单日期图像-掩码对；它们通常局限于特定区域而非全球覆盖，或对公众不可访问。为了解决这些限制，MANGO数据集被创建，它是一个全球性的、单日期配对红树林分割数据集。该数据集包含高分辨率卫星图像与对应的红树林掩码，覆盖了全球主要红树林分布区域。数据采集过程结合了多种卫星影像源，并利用先进的地理信息系统技术进行预处理和标注。MANGO数据集的构建旨在为红树林监测和保护提供一个标准化、可访问的基准。其核心特点包括：1. 全球覆盖：数据点分布在亚洲、非洲、北美洲、南美洲和大洋洲等多个大洲，涵盖了不同气候带和生态环境下的红树林类型。2. 单日期配对：每个样本都包含特定日期的卫星图像及其对应的红树林分割掩码，这对于训练和评估时间敏感的深度学习模型至关重要，能够更准确地捕捉红树林生态系统的动态变化。3. 高分辨率：采用高分辨率卫星影像，确保了红树林边界和内部结构特征的精细化识别。4. 统一标注标准：所有掩码均遵循统一的标注规范，保证了数据集的内部一致性和可用性。5. 可访问性：MANGO数据集将公开发布，以促进全球研究人员在红树林监测领域的合作与创新。该数据集的发布有望显著推动基于深度学习的红树林制图、生物量估算、碳储量评估以及病虫害监测等应用的发展，为红树林生态系统的可持续管理提供数据支撑。
📄 English Summary
MANGO: A Global Single-Date Paired Dataset for Mangrove Segmentation
Mangroves are crucial for climate change mitigation, necessitating reliable monitoring for effective conservation. Deep learning has emerged as a powerful tool for mangrove detection, but its progress is hampered by limitations in existing datasets. Specifically, many resources offer only annual map products without curated single-date image-mask pairs, are restricted to specific regions rather than global coverage, or remain inaccessible to the public. To address these limitations, the MANGO dataset was developed as a global, single-date paired dataset for mangrove segmentation. This dataset comprises high-resolution satellite imagery paired with corresponding mangrove masks, covering major mangrove distribution areas worldwide. The data acquisition process integrates multiple satellite image sources and leverages advanced Geographic Information System (GIS) techniques for preprocessing and annotation. The construction of the MANGO dataset aims to provide a standardized and accessible benchmark for mangrove monitoring and conservation. Its key features include: 1. Global Coverage: Data points are distributed across various continents, including Asia, Africa, North America, South America, and Oceania, encompassing diverse mangrove types under different climate zones and ecological environments. 2. Single-Date Pairing: Each sample includes a satellite image from a specific date and its corresponding mangrove segmentation mask. This feature is crucial for training and evaluating time-sensitive deep learning models, enabling more accurate capture of dynamic changes within mangrove ecosystems. 3. High Resolution: High-resolution satellite imagery is utilized, ensuring detailed identification of mangrove boundaries and internal structural characteristics. 4. Uniform Annotation Standards: All masks adhere to consistent annotation guidelines, guaranteeing internal consistency and usability of the dataset. 5. Accessibility: The MANGO dataset will be publicly released to foster global collaboration and innovation in mangrove monitoring research. The release of this dataset is expected to significantly advance deep learning-based applications such as mangrove mapping, biomass estimation, carbon stock assessment, and pest and disease monitoring, thereby providing data support for the sustainable management of mangrove ecosystems.